人体摄入放射性核素后会形成体内放射性污染，而全身计数器正是用于定量评价这种污染的重要设备。NaI型全身计数器是目前该设备中最常用的类型，但其获取的γ能谱较为复杂，使用现有方法难以准确分析。为此，本文开展了相关研究，对现有方法进行改进的同时还发展出了两种新方法，具体过程和成果如下：1）首先前往了国内10家核电厂，进行了现场测量和能谱收集，此外，还完成了对一台设备的精细数字化建模。基于收集的实测能谱，研究了NaI型全身计数器γ能谱的特征，得到了低能散射峰的特点和成因以及3种核电厂典型内污染核素57Co、137Cs、60Co的γ能谱全能峰基底形态。2）改进了γ能谱全能峰基底的计算方法，不再采用以往的线性或多项式拟合方法，而是提出了以余误差函数为模型的拟合法及以递推关系式为基础的非拟合法。研究表明，拟合法适用于单峰情况，非拟合法适用于重峰以及基底计数较高的情况，且效果都显著优于以往方法。不仅如此，利用上述成果并结合已有的全能峰分析算法，还开发了国内首个NaI型全身计数器专用的γ能谱分析软件NASAS，实验发现，与Canberra公司商用软件Genie2000相比，NASAS对低能核素的分析计算结果更好。3）开展了利用响应矩阵进行γ能谱分析的方法研究，建立了γ射线从发出到被探测的响应矩阵，并首次考虑了能谱的稀疏性，利用稀疏算法进行能谱反解。结果显示，响应矩阵法对核素活度的估算值与实际值偏差小于6%，显著优于全能峰分析法。4）在定量分析之外，还基于机器学习算法研究了NaI型全身计数器γ能谱的异常定性识别方法。该部分以现场收集的大量实测能谱为学习对象，建立了以主成分分析进行特征提取，以Mahalanobis距离作为分类器的完整方法流程。应用结果证明，该方法在识别异常上较前述几种定量分析法更灵敏，且误识别情况更少。 本文的工作使得提高NaI型全身计数器测量准确性成为可能，进而为确保该类型设备在核工业现场、核事故应急情况下使用的可靠性提供了支持。
Whole Body Counter (WBC) is an important equipment to quantitatively evaluate the in-vivo pollution of human body after the intake of radionuclides. NaI-type WBC is the most commonly used WBC, however, its measuring result—the γ spectrum—is usually very complex, and difficult to be accurately analyzed by existing methods. To this end, this article carried out the relevant research, not only to improve the existing methods, but also developed two new methods. The specific process and results are as follows: 1) Carried out on-site measurement andγspectrum collection in 10 domestic nuclear power plants, and also completed a fine digital modeling of one WBC. Based on the collected measured γ spectrum, the characteristics of γ-ray spectrum of NaI-type WBC are studied which discovered the forming reason of low energy scattering peak and the shape of the continuum under full-energy peaks generated by 3 typical inner-contamination nuclides:57Co、137Cs、60Co. 2) The method of calculating the continuum under full-energy peaks is improved. the conventional linear or polynomial fitting method is no longer used. the fitting method based on Residual Error function and the non-fitting method based on recursive formula are proposed. The results show that the fitting method is applicable to the single peak condition and the non - fitting method is suitable for the case of multiplets and high continuum count, and the performance is obviously better than the previous method. Not only that, with this fruit and the existing full-energy peak analysis algorithms, the first domestic NaI-type WBC dedicated gamma spectrum analysis software NASAS is also developed, and the experiment found that, compared with Canberra's commercial software Genie2000, NASAS’s analysis result for low energy nuclides are better. 3) The study of γ spectrum analysis method using the response matrix is carried out. The response matrix from γ-ray being released to be detected is established, and for the first time, the sparseness of the energy spectrum is considered, so the inverse of the spectrum is solved by sparse algorithm. The results show that the bias of the estimated nuclide activity value given by response matrix method is less than 6% compared to the actual value, which is significantly better than the full-energy peak analysis method. 4) In addition to quantitative analysis, the anomalous qualitative identification method for NaI type WBC γ spectrum is also studied based on the machine learning algorithm. In this part, a large number of measured energy spectrum collected in the field is used as the learning object, and the method, which uses Principle Component Analysis for feature extraction with the Mahalanobis distance as the classifier, is established. The results show that this method is more sensitive with less misidentification compared to these quantitative analysis methods above. The work of this paper makes it possible to improve the accuracy of the NaI-type WBC measurement, and thus to ensure the reliability of this type of equipment to be used in nuclear industry as well as in nuclear accident emergency situations.
